An Interpretable Convolutional Neural Network Framework for Analyzing Molecular Dynamics Trajectories: a Case Study on Functional States for G-Protein-Coupled Receptors

Molecular dynamics (MD) simulations have made greatcontribution to revealing structural and functional mechanisms formany biomolecular systems. However, how to identify functional statesand important residues from vast conformation space generated by MDremains challenging; thus an intelligent navigation is highly desired.Despite intelligent advantages of deep learning exhibited in analyzingMD trajectory, its black-box nature limits its application. To addressthis problem, we explore an interpretable convolutional neural network(CNN)-based deep learning framework to automatically identifydiverse active states from the MD trajectory for G-protein-coupledreceptors (GPCRs), named the ICNNMD model. To avoid theinformation loss in representing the conformation structure, the pixelrepresentation is introduced, and then the CNN module is constructedto efficiently extract features followed by a fully connected neuralnetwork to realize the classification task. More importantly, we design a local interpretable model-agnostic explanation interpreter forthe classification result by local approximation with a linear model, through which important residues underlying distinct active statescan be quickly identified. Our model showcases higher than 99% classification accuracy for three important GPCR systems withdiverse active states. Notably, some important residues in regulating different biased activities are successfully identified, which arebeneficial to elucidating diverse activation mechanisms for GPCRs. Our model can also serve as a general tool to analyze MDtrajectory for other biomolecular systems. All source codes are freely available athttps://github.com/Jane-Liu97/ICNNMDforaiding MD studies